Electrical theft is a global issue that harms both utility providers and electrical users. It destabilizes utility companies' economic development, creates electric dangers, and raises energy costs for customers. The development of smart grids is significant in power theft detection because they generate huge amounts of data, including consumer usage data, which may be used to detect electricity theft using machine learning and deep learning algorithms. This study introduces a deep neural network-based classification method for detecting theft that uses a lot of data in the time and frequency domains. Data interpolation and synthetic data creation procedures are applied to address dataset shortcomings such as missing values and class imbalance. The competitiveness of the proposed strategy is demonstrated in comparison with other methods evaluated on the same dataset. During validation, the approach achieves a 90% area under the curve (ROC), which is 1% higher than the best-performing DNN currently available, and an accuracy of 94.48%, the second-highest on the benchmark.